With perennial crops, it is necessary to select individual trees when applying certain selection strategies. Choosing individuals and multiplying them by vegetative propagation is an essential step in selection of new clone varieties. The selection of new parental genotypes within segregating progenies is a basic feature of recurrent selection schemes. For both these objectives, it is necessary to have reliable methods for choosing individuals in experimental plots. Various methods are proposed here for improving the selection efficiency for quantitative traits of individuals in segregating progenies in cocoa variety trials. Firstly, statistical methods are presented that make it possible to fine-tune the choice of trees for selection: "combined individual-family" selection, "smoothing methods" derived from spatial statistics and "longitudinal data analysis". Secondly, methods are proposed to improve experimental designs by increasing the number of trees per genotype: use of "true twins" and rapid vegetative propagation of individual seedlings. These different methods are discussed and ways of making the most effective choice of cocoa trees in experimental plots are proposed.